🔥Telegram İfşa

🚀 Excited to share our latest work at #ICLR2024! 📄 "Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond", conducted during my internship at Amazon! 🌐Leveraging multi-modal data for various personalized tasks.

🚀Excited to share "Language Models as Semantic Indexers" is accepted to ICML 2024! ⭐️We propose to learn document semantic IDs with large language models in a self-supervised fashion. ⭐️The learned semantic IDs can benefit LLM generative recommendation and retrieval. #LLM #IR

Join us next Friday, June 28th, for our last SEA meet-up before the summer break! 🎉 @snbruch (Pinecone) and Hansi Zeng (UMass Amherst) will discuss recent innovations in search indexing for dense retrieval and in scaling generative IR. Sign up here 👉 meetup.com/de-DE/sea-sear…

Excited to be in DC for SIGIR! My first time for an in-person conference! I'll present my paper, "Planning-Ahead in Generative Retrieval" tomorrow at 3:05 pm (Monday). Don't miss the M2.1 GenIR and future of LLMs for search session. Looking forward to chatting with IR folks!

Not arrived at Vienna!😅 Sad that cannot make it this time. #ICML2024 However, if you are interested in 𝗟𝗟𝗠 𝗮𝗻𝗱 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹, come and check out our paper tomorrow afternoow at Hall C 4-9 #404! paper link: arxiv.org/pdf/2310.07815

🔥 Unleashing the Power of Inference Scaling in Long-Context RAG 🔥 Excited to share our latest research Google DeepMind "Inference Scaling for Long-Context Retrieval Augmented Generation (RAG)" 🎓, in collaboration with Honglei Zhuang Aijun Bai Kai Hui Rolf Jagerman Hansi Zeng

📢 An excellent opportunity for PhD students in IR and NLP: The Center for Intelligent Information Retrieval (CIIR) at the UMass Amherst is initiating an exciting Research Internship program for Summer 2025. See the thread for more info. 👇 #SIGIR #NLProc

LLM Alignment as Retriever Optimization: An Information Retrieval Perspective arxiv.org/abs/2502.03699 We introduce a comprehensive framework that connects LLM alignment techniques with the established IR principles, providing a new perspective on LLM alignment.

🎉I'm happy to announce my first PhD paper: "Hypencoder: Hypernetworks for Information Retrieval" 🎉 We investigate a way to model relevance beyond inner-products. Instead of using a query vector, we use a query-specific neural net produced by a hypernetwork encoder (Hypencoder)

Scaling Sparse and Dense Retrieval in Decoder-Only LLMs Hansi Zeng et al investigate how different retrieval paradigms scale with larger models, showing sparse retrieval consistently outperforms dense retrieval while demonstrating better generalization 📝arxiv.org/abs/2502.15526

🚀 Introducing 𝗦𝗲𝗮𝗿𝗰𝗵-𝗥𝟭 – the first 𝗿𝗲𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻 𝗼𝗳 𝗗𝗲𝗲𝗽𝘀𝗲𝗲𝗸-𝗥𝟭 (𝘇𝗲𝗿𝗼) for training reasoning and search-augmented LLM agents with reinforcement learning! This is a step towards training an 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 𝗢𝗽𝗲𝗻𝗔𝗜 “𝗗𝗲𝗲𝗽

🚀 Excited to announce that our paper 𝐒𝐞𝐚𝐫𝐜𝐡-𝐑𝟏 is now live! 📄 We introduce an RL framework (an extension of 𝐃𝐞𝐞𝐩𝐬𝐞𝐞𝐤-𝐑𝟏) for training reasoning-and-retrieval interleaved LLMs. We’re also open-sourcing all resources—models, data, and more! 📜 Paper:

Can LLMs learn to search and reason interleavedly for complex QA tasks without predefined rules? Yes! 🚀 Check out Search-R1, our framework that leverages the DeepSeek-R1-style RL approach to achieve this.

I’m thrilled to share that two of my papers were accepted to SIGIR 2025: “Hypencoder: Hypernetworks for Information Retrieval” and “Scaling Sparse and Dense Retrieval in Decoder-Only LLMs” with Hansi Zeng Hamed Zamani

🚨 Big updates to 𝗦𝗲𝗮𝗿𝗰𝗵-𝗥𝟭! 🚀 🧠 Now supports 𝗺𝘂𝗹𝘁𝗶-𝗻𝗼𝗱𝗲 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 — train 32B+ LLMs with search + reasoning: github.com/PeterGriffinJi… 🔍 Added support for 𝗹𝗼𝗰𝗮𝗹 𝘀𝗽𝗮𝗿𝘀𝗲/𝗱𝗲𝗻𝘀𝗲 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗲𝗿𝘀 & 𝗼𝗻𝗹𝗶𝗻𝗲 𝘀𝗲𝗮𝗿𝗰𝗵